Assessment of Disease Activity Using a Whole-body MRI Derived Radiological Activity Index in Chronic Nonbacterial Osteomyelitis


 Background. Based on the recently developed ChRonic nonbacterial Osteomyelitis MRI Scoring tool (CROMRIS), we developed a radiological activity index (RAI-CROMRIS) to obtain a quantification of the overall bone involvement in individual patients.Methods. Whole Body Magnetic Resonance Imaging (WB-MRI) images were scored according to parameters included in the RAI-CROMRIS: bone marrow hyperintensity, signal extension, soft tissue/periosteal hyperintensity, bony expansion, vertebral collapse. These parameters were evaluated for each bone unit yielding a score from 0 to 7 and summed up as RAI-CROMRIS including all bone units. We assessed clinical disease activity using a physician global assessment (PGA) and radiological findings in 76 treatment-naïve patients; 46 of 76 were evaluated at 6 and 12 months after initial WB-MRI. Quantitative variables were compared using the Mann-Whitney U test for unmatched groups and the Wilcoxon signed-rank test for paired groups. Correlation was evaluated using Spearman's rank coefficient (rs).Results. There was a significant correlation between RAI-CROMRIS and PGA (rs=0.32; p=0.0055), between RAI-CROMRIS and presence of elevated erythrocyte sedimentation rate (p=0.013) and C-reactive protein (p=0.0001) at baseline. The RAI-CROMRIS decreased from a median of 17 at baseline to 12 at 6 months (p=0.004) and remained stable (median 11) at 12 months. A correlation between the RAI-CROMRIS and the PGA was observed at baseline (rs=0.41; p=0.004) and during follow up at 6 months (rs=0.33; p=0.025) and 12 months (rs=0.38; p=0.010). The baseline RAI-CROMRIS (median 20) was significantly higher in patients who subsequently received bisphosphonates than in patients who received other treatments (median 12) and decreased significantly after bisphosphonates (p=0.008).Conclusions. The RAI-CROMRIS was correlated with clinical and laboratory measures of disease activity showing significant short-term changes following treatment with bisphosphonates. This tool could be used in clinical practice and clinical trials after validation.


Background
Chronic non-bacterial osteomyelitis (CNO) is an autoin ammatory bone disease, characterized by sterile bone lesions due to an abnormal regulation of the innate immune response resulting in osteoclast differentiation and activation, osteolysis and bone remodeling [1,2]. It was rst described in 1972 as an unusual form of subacute and chronic symmetrical osteomyelitis [3]. Nowadays the term chronic nonbacterial osteomyelitis is preferred [4,5].
Incidence rate of CNO varies 1-4 per million children from different reports [6] and an incidence of approximately 1:1,000,000 [7] or 2-5% of all osteomyelitis cases [8]. However, this may be underestimated due to low awareness. The average age of onset ranges between 7 and 12 years [4,9] with a preference for the female gender [9,10]. The disease is characterized by bone in ammation. The etiopathogenesis still remains unclear. It is believed to be polygenic and multifactorial leading to an imbalanced expression of anti-in ammatory/immunomodulating (IL-10 and IL-19) and proin ammatory (IL-1β, IL-6, TNF-α, IL-20) cytokines [2,11,12].
Clinical presentation and severity of CNO vary signi cantly from asymptomatic subjects at one end to patients with multiple recurrent lesions with signi cant functional impact at the other end [13,14]. The cardinal feature is bone pain. Onset is often insidious and patients may or may not experience concomitant systemic symptoms including low-grade fever and generalized malaise [9,15]. About 30% of patients had unifocal disease at diagnosis but most of them developed new lesions during the follow-up period [16]. Only 25-40% of individuals have symmetrical bone involvement [4]. Finally, a subset of CNO patients exhibit extra-osseous in ammatory organ involvement, including psoriasis and palmoplantar pustulosis (~ 8%), severe acne (~ 10%) and in ammatory bowel disease (~ 10%) [13,14,16,17].
The diagnosis of CNO still remains a diagnosis of exclusion and a bone biopsy is often advocated to con rm the diseases. Imaging is becoming increasingly important in the diagnosis and follow-up of these patients [18,19]. In ammatory lesions in CNO may be detected by conventional radiographs with low sensitivity [20]. Computed tomography (CT) may show lytic lesion and sclerosis at a higher sensitivity than radiograph in CNO. Its use is limited due to the high radiation dose. Bone scintigraphy and positron emission tomography (PET)-CT are sometimes used to offer a whole-body level assessment in order to identify lesions simultaneously [21]. However, both are radiation-based techniques. Whole-body magnetic resonance imaging (WB-MRI) represents the current gold standard in the diagnosis and management of CNO patients. It allows the assessment of multiple affected sites simultaneously through typical ndings of marrow edema (hypointense on T1-weighted images and hyperintense on T2-weighted images, or short tau inverse recovery [STIR] during the active phase of the disease) [22,23]. MRI can also demonstrate associated periosteal reaction, soft tissue in ammation and transphyseal disease. In addition, clinically asymptomatic lesions, also known as "silent" lesions, can often be detected by MRI.
WB-MRI is radiation-free and well suited for surveillance and monitoring of therapeutic response; however it requires sedation in younger children.
Recently a standardized evaluation tool of WB-MRI for CNO has been developed by an international group of pediatric radiologists: the ChRonic nonbacterial Osteomyelitis MRI Scoring (CROMRIS) [24]. Based on this tool, we developed a radiological activity index (RAI-CROMRIS) to obtain a quantitative measure of bone involvement at multiple sites in individual patients.

Patients
We retrospectively analyzed 76 patients with CNO followed at the division of Rheumatology of "Ospedale Pediatrico Bambino Gesù" from April 2012 to September 2019. Inclusion criteria were: ful lling the CNO criteria proposed by Jansson et al [25] and WB-MRI performed at baseline, before starting any speci c treatment. Complete clinical data were collected as described below.
Clinical disease activity was evaluated using a physician's global assessment (PGA), based on a 5-point scoring system (inactive, minimal, mild, moderate, severe disease) that included fever, increase of in ammatory markers, presence of pain and/or functional impairment and physician visual analogue scale (VAS) modi ed from what described in Pardeo et al. [26] ( Table 1). The intensity of pain was assessed by the clinician on a VAS ranging from 0 to 10, with 0 as no pain and 10 as maximum pain. Normal ranges used for laboratory data were as follows: C-reactive protein (CRP) < 0.5 mg/dl and erythrocyte sedimentation rate (ESR) ≤ 15 mm/h. These clinical data were collected for all patients with baseline WB-MRI as well as those who had WB-MRI at follow up. To evaluate clinical disease activity we used a PGA scoring system based on fever, presence of pain and/or functional impairment, increase of in ammatory markers and physician VAS scale. Total score: 0 = inactive, 1 = minimal, 2 = mild, 3 = moderate and 4 = severe.

Whole body MRI protocol and scoring
The WB-MRI protocol is available in Additional was also available at 6 months (MRI T6) and at 12 months (MRI T12) for 46 patients after starting therapy. MRI images were reviewed on a dedicated PACS workstation by two radiologists in consensus reading, with respectively 2 years and 10 years of experience in pediatric radiology. Readers were blinded to patients clinical, anamnestic and laboratory data. WB-MRI images were assessed using the CROMRIS as published by Zhao et al. [24] (Additional Table A 2 ). The CROMRIS atlas included evaluation of 20 bone units (the entire skeleton was divided into long bones, complex bones, hand/foot and spine) using 4 different variables: bone marrow hyperintensity, size of signal intensity within each unit/segment graded using relative measurement (small de ned as < 25% of estimated volume, medium as 25-50% and large > 50% of estimated volume), hyperintensity of surrounding tissue and bony expansion. According to CROMRIS, we decided to include the following parameters for only active CNO lesions: 1) presence of bone marrow hyperintensity on STIR images (score 0-1), 2) presence of signal extension (score 0-3), 3) presence of soft tissue/periosteal hyperintensity (score 0-1), 4) presence of bony expansion (score 0-1), 5) presence of vertebral collapse (score 0-1) in our RAI-CROMRIS. These parameters were evaluated for each bone unit yielding a score from 0 to 7 ( Table 2). The minimum score for each identi ed bone unit is 1 as the presence of bone marrow hyperintensity. The total RAI-CROMRIS was obtained by adding up the scores of all bone units. Joint effusion (synovitis), and measures of damage/complications, such as non-CNO bony abnormalities, limb hypertrophy, kyphosis, bony expansion without bone edema and pathological fractures were excluded from the index developed here.
The RAI-CROMRIS was calculated for all 76 patients who had a WB-MRI at baseline and for all the 46 patients who had WB-MRI at 6 and 12 months. All the patients had complete clinical and laboratory data. The parameters included and score assigned to each variable for active CNO lesions. Adding up all parameters a maximum score of seven was obtained for each bone.
*In long bones the extension of bone edema < 25% was de ned if one part out of ve was involved, between 25% and 50% if two parts out of ve were involved and of > 50% if three or more parts were involved. In complex bone and on the spine extension was de ned as in CROMRIS scoring with 1 point an extension < 25%, 2 points an extension between 25% and 50% and 3 points with an extension > 50%.

Statistical Analysis
Qualitative variables were expressed as absolute frequency and percentage. Proportions were compared by Chi-square test or Fisher's exact test, as appropriate. Quantitative variables, reported as medians and interquartile ranges (IQR: 1st -3rd quartile), were analyzed using the Mann-Whitney U test for unmatched groups and the Wilcoxon signed-rank test for paired groups.
Correlation between quantitative variables was evaluated using Spearman's rank coe cient (r s ) and the strength of the correlation were classi ed as follows: r s <0.40 = weak, 0.40-0.59 = moderate, 0.60-0.79 = strong and ≥ 0.80 = very strong [27].
All statistical tests were two-sided; a p value < 0.05 was considered as statistically signi cant. The analyses and graphs were performed using Stata 15.1 software (StataCorp LLC, College Station, Texas USA, 2017).

Results
Demographic data, clinical and radiological ndings at baseline The demographic and baseline characteristics of patients are shown in Table 3. We evaluated 76 patients   In the group of patients who did not receive bisphosphonates therapy (35/76): 16 patients have been treated with only nonsteroidal anti-in ammatory drugs (NSAIDs), 15 patients received SSZ and/or MTX, 1 patient received biologic agents, 3 patients received SSZ and/or MTX with concomitant biological therapy.
At baseline, we found a signi cant, but weak, correlation between the RAI-CROMRIS and the PGA (r s =0.32; p = 0.0055) (Fig. 1A). We also found a signi cant association of the RAI-CROMRIS with the presence of functional impairment, and with ESR and CRP above the upper limits of normal (Fig. 1B, C and D The RAI-CROMRIS decreased from a median of 17 at baseline (IQR 12-26) to a median of 12 at 6 months (IQR 6-20; p = 0.004) and remained stable at a median of 11 at 12 months (IQR 4-20; T12 vs baseline p = 0.003 and T12 vs T6 p = 0.52) ( Fig. 2A). The distribution of the PGA scores and the relation with the RAI-CROMRIS at different time points is shown in Fig. 2B. A signi cant correlation between the RAI-CROMRIS and the PGA was observed at baseline and during follow up with a moderate correlation at baseline (r s =0.41, p = 0.004) and a weak correlation at 6 months (r s =0.33, p = 0.025) and at 12 months (r s =0.38, p = 0.010). We also found that the changes in the RAI-CROMRIS from baseline to 12 months were signi cantly associated with the changes in the PGA (r s= 0.43, p = 0.003).

RAI-CROMRIS at baseline and at 6 months in patients treated with bisphosphonates
As the optimal treatment of CNO has not yet been established, the therapeutic choices for CNO patients are driven by the treating physician's judgement. Bisphosphonates are usually proposed in current treatment plans for more severe patients [28]. We retrospectively compared RAI-CROMRIS values at baseline and changes at 6 months in patients who received bisphosphonates (n = 27) with those who did not (n = 19). We found that patients who subsequently received bisphosphonates had higher baseline RAI-CROMRIS compared to patients who received other treatments (median 20, IQR 13-42 vs. median 12, IQR 8-18; p = 0.008).

Discussion
Whole-body MRI (WB-MRI) is one of the mainstays in the diagnosis and management of CNO patients. It is becoming the imaging modality of choice because of its sensitivity in detecting lesions and their characteristic pattern, as well as clinically silent lesions, without ionizing radiation [29]. However, the issue of appropriate interpretation, particularly in children growing bones, has led a number of investigators to propose standardized MRI protocols and scoring systems with the aims to improve reproducibility and inter-examiner variability and, ultimately, diagnosis and staging of the disease [30,31]. Some of the available scoring systems have shown merits, but also some limitations and they have not been validated. One is an MRI score system for the osteitis lesions used in patients who met the criteria for SAPHO syndrome (Synovitis, Acne, Pustulosis, Hyperostosis, Osteitis syndrome). This score ranges from 0 to 2 with score 0 de ned as no bone marrow edema, osteal erosions or synovitis (with or without joint effusion); score 1 de ned as the presence of one of these ndings; and score 2 de ned as the presence of 2 or more ndings. In case of more than one osteitis lesion, the lesion with the highest score was used as reference for that patient [30,32]. In the RINBO index, points are assigned for each of four parameters of interest: the number of radiologic active lesions clustered into 3 categories as unifocal, paucifocal [2-4 lesions] and multifocal [5 or more lesions], the diameter of the largest lesion, the presence of extra medullary involvement, the presence of spinal involvement [33]. Differently from the previous score, the RINBO generates a total score summing the score of the lesions. The main limitation is that the measurement of the size of lesions is not reported as relative to the size of each bone, which is more appropriate, particularly, in children.
We chose to use as a basis of a global quantitative radiological assessment the CROMRIS, a standardized grading of the different features of CNO lesions, irrespective of the bone involved, recently designed by Zhao et al [24]. It is based on CNO features previously used to score region-speci c MRI [31], but with modi cations developed, through a consensus process, by 11 pediatric experienced radiologists from 7 different centers and 2 continents (North America and Europe). These authors produced, after a literature review, a complete atlas that consists of detailed de nitions and grading of signal intensity, size of signal abnormality within bone marrow and surrounding tissue, physis damage and vertebral compression for each individual bone unit. This method showed excellent reliability and agreement in each category of bones and moderate to substantial reliability and agreement in readings from individual bones. This score was designed to be easily applied in settings with non-experienced radiologists. Therefore, the CROMRIS represents an excellent standardized semi-quantitative scoring system. A total score re ecting the radiological activity of each individual patient has not been proposed. A total score taking all lesions into account is needed because the uctuating course of CNO implies that new lesions may occur while others are vanishing. Hence, we chose to integrate scores from each bone unit in a comprehensive RAI-CROMRIS that could measure the "entire radiological burden" of the disease at a whole-body level.
Few studies with small cohorts addressed the direct correlation between radiological and clinical ndings in CNO patients. A comparison of clinical and radiological ndings was previously performed in small cohort without a quantitative radiological score [34,35]. More recently, the RINBO score was evaluated in 40 patients and found to have a fair agreement with the clinical evaluation of each bone site; however, patients' medication and follow-up scans were not taken into account [33]. In our study we evaluated the correlation of the RAI-CROMRIS with disease activity in 76 patients before starting treatment and in 46 patients also at two subsequent time points (6 and 12 months). We found a signi cant association of the RAI-CROMRIS with disease activity parameters including the PGA, presence of functional impairment and increased in ammatory markers in patients at baseline. These observations suggest that the burden of radiological activity, as measured by the RAI-CROMRIS, re ects the overall degree of clinical activity of each individual patient. Examining 46 patients with WB-MRI at baseline and at 6 and 12 months, we found a signi cant decrease in the RAI-CROMRIS overtime. Importantly, a correlation between the RAI-CROMRIS and the PGA was observed, not only at baseline, but also during follow up at 6 and 12 months, at times when the disease activity was signi cantly lower, suggesting that this score may be able to re ect minor changes in disease activity.
Several studies have shown the effectiveness of bisphosphonates as a treatment option for patients with signi cant disease burden, with physical limitations or active vertebral lesions or in those who have persistently active symptoms and abnormal MRI ndings who have failed others treatment (NSAID, DMARDS, biologic agents) [28,31,36,37]. This is re ected by the recently developed consensus treatment plans [28]. The patients whom we chose to treat with bisphosphonates, not only had a more severe disease (as shown by more frequent functional impairment and more frequent vertebral involvement), but also had a signi cantly higher baseline RAI-CROMRIS compared to patients who received other treatments. After 6 months of treatment, the improvement in disease activity was associated with a signi cant decrease in the RAI-CROMRIS. The decrease in the score was particularly evident for extension of bone hyperintensity, as well as for soft tissue hyperintensity. As hyperintensity is indicative of in ammatory edema, this observation is consistent with the hypothesis of an antiin ammatory effects of targeting osteoclasts in CNO.
There are some limitations of our study. Even if our series is large, the monocentric nature of the study, its observational design, the heterogeneity of the treatments used and the rather long time interval in which the patients were enrolled require caution in drawing de nitive conclusions. Although we have used a comprehensive PGA including different variables, such as physician reported outcome, patient reported outcome and increase in acute phase markers, this is not validated. Indeed, there is not yet an internationally validated tool for measuring disease activity in CNO. It should also be noted that bone hyperintensity on STIR sequences may be present in healthy children during growth and, therefore differentiating this physiological hyperintensity from the pathological in ammatory oedema of CNO may be challenging [38]. Bone-marrow-edema like changes in the hand skeleton are reported in healthy children aged 5-16 years in the hand skeleton [39,40], as well as in feet [41] and pelvis [42]. The low prevalence of bone hyperintensity within these sites in our cohort and the relevant change after treatment argue against over-reading. Furthermore, a potential limitation of the CROMRIS is that the parameters, except extension of hyperintensity in each bone unit, have the same scoring range, which may have diluted an important variable. Also all bone units were considered equally, though hands bones and foot bones were grouped together and assigned the same unbiased weight. Whether these issues are relevant in the evaluation of the total radiological burden of disease and its relation to clinical disease activity remains to be established. Our study provides follow-up information up to 1 year, which is informative on the short term, but longer-term follow-up is needed in a chronic relapsing disease such as CNO.
Radiological follow-up studies are lacking. In the only available follow-up study WB-MRI performed more than 10 years after disease onset, revealed radiologically active lesions in more than 50% of patients even in long-term clinical remission [35].

Conclusions
In conclusion, our results show that the RAI-CROMRIS provides a measure of the overall radiological burden of disease in individual CNO patients and that it is related to clinical and laboratory measures of disease activity, as well as shows signi cant short-term changes following treatment with bisphosphonates. Larger studies, ideally multicentric, are needed to evaluate the usefulness of the RAI-CROMRIS to stratify patients with CNO based on the severity of the disease, and consequently to guide the most appropriate therapeutic choice, and to evaluate its sensitivity to change following treatments. This radiological score could be included in clinical trials aimed to evaluate the e cacy of targeted treatments in CNO.

Declarations
Ethics approval and consent to participate: The study was approved from the ethics committee of the "Ospedale Pediatrico Bambino Gesù" with the ethics approval number 2144/2020.

Consent for publication:
not applicable.
Availability of data and materials: the datasets used and analysed during the current study are available from the corresponding author on reasonable request.
Competing interest: the authors declare that they have no competing interests.
Funding Statement: no speci c funding was received from any bodies in the public, commercial or not-for-pro t sectors to carry out the work described in this article.